This paper demonstrates a practical application of an established approach for analyzing the CEER database, Lombardy Region’s Energy Performance Certificates (EPCs) repository. The objective is to develop building archetypes using an advanced unsupervised clustering technique. The comprehensive methodology processes and refines CEER’s data specific to Lombardy, incorporating critical parameter selection, rigorous quality checks, and effective data clustering. Key parameters such as building category, climatic region, year of construction, and various thermal properties have been meticulously analyzed. The study aims to enhance the accuracy and efficiency of Urban Building Energy Models (UBEMs) by developing archetypes that represent typical buildings with similar characteristics. Different algorithms are compared to optimize clustering, including K-means and K-means++, evaluating their performance using the Davies-Bouldin index. The optimized methodology demonstrates improved clustering quality and faster convergence, significantly reducing computational time while ensuring robustness and scalability, providing a replicable and scalable analytical methodology for building characterization. The resulting archetypes provide a simplified yet accurate representation of building energy characteristics, reducing the complexity and preparation time required for UBEM simulations. This contributes to lowering uncertainties in energy modeling, offering a replicable and scalable analytical framework that public authorities and modelers can utilize to enhance the use of UBEM and building energy efficiency in general. The findings underscore the potential of leveraging EPCs to create building archetypes, ultimately supporting sustainable urban planning and energy management initiatives.